Résumé
Effects of variation in spawning stock and sea temperature on long-term temporal patterns in recruitment dynamics of 38 commercially harvested fish stocks in the northern North Atlantic were studied. Different statistical models were explored within a Ricker stock-recruitment framework. This includes, in order of complexity, adding a linear temperature term, a nonlinear (smooth) temperature effect, and non-stationarities (trends in intercept or in temperature effect) and finally allowing for a stepwise change (a threshold). The different models were compared in a uniform approach using Akaike’s information criterion corrected for small sample size as the model selection criterion. The relationship between recruitment, spawning stock biomass and temperature varied over time. The most frequent alteration in the non-stationary linear models was, for 14 stocks, in the intercept in recruitment success, suggesting a change in pre-recruit mortality over time. Threshold models performed better than the best linear or nonlinear stationary models for 27 of the stocks, suggesting that abrupt changes (maybe even regime shifts) are common. For half of the stocks studied, the temperature effect was statistically significant when added to the model of the relationship between recruitment success and spawning stock biomass. This includes all 6 of the herring stocks studied, with a positive effect for cold-water stocks and negative effect for stocks in the more temperate southern areas. For the 4 plaice stocks analysed, all located towards the centre of the overall distribution range of plaice, a tendency toward recruitment being favoured by lower temperatures was found.